Experimental Establishment of Phase Diagrams Guided by Uncertainty Sampling: An Application to the Deposition of Zn–Sn–P Films by Molecular Beam Epitaxy
Ryoji Katsube, Kei Terayama, Ryo Tamura, Yoshitaro Nose
Abstract
The construction of phase diagrams is an essential step in the development of materials and devices, but this process requires considerable time and the efforts of well-established researchers. Recently, a machine learning method based on uncertainty sampling, phase diagram construction (PDC), was developed in order to minimize the number of required experiments. In the present study, we, for the first time, demonstrate PDC-assisted experimental determination of an uninvestigated phase diagram for the deposition of Zn–Sn–P films by molecular beam epitaxy (MBE). Interestingly, without any background knowledge of materials science, an undetected phase in the initial dataset was detected after only 4 cycles of PDC-assisted experiments, and the phase diagram was sufficiently refined after only 11 cycles. Based on the results and thermodynamic discussions, we confirm the effectiveness of PDC on the construction of phase diagrams for MBE process. The demonstration in this study shows the machine-learning method, PDC, will accelerate discovery and development of materials; furthermore, we discuss how to use PDC more effectively, considering the objectives involved in the construction of phase diagrams.